The statements in this section merely provide background information related to the present disclosure and do not necessarily constitute prior art.
Random numbers are usually generated by using a signal source with a noise characteristic or a metastable characteristic for sampling the signal source outputs and comparing the same with a reference value. It is noted that all output values of the random number are supposed to have no correlation, and it is necessary to have a uniform frequency of occurrence of all n values that make up the n-ary random numbers.
However, it is also noted that when using thermal noise as the signal source or using a metastable state, considerable correlation occurs between the signals before and after the generated random number, wherein the proportion of a particular value is far greater than those of other values, which is called biasing.
FIG. 1 is a diagram illustrating biasing present in the resulting sample values from sampling a signal source with deviation.
When a deviation occurs in the signal source due to changes in the external environment or for other reasons, its resulting sample values inherit the biasing that occurs due to a high frequency of occurrence of a particular value. For example, section B of the resulting sample in FIG. 1 shows 1s more frequently than 0s, which is due to the signal source having the deviation in that time zone.
Since such biasing lowers the utility of generated random numbers, there are various methods, attempting to overcome the biasing issue. It is noted that the typical solution to the biasing issue is analyzing sample signals and monitoring whether values that make up the random numbers are generated in a certain ratio for performing appropriate control and signal processing.
For example, a structure for generating a random number using noise as a signal source is configured in such a way that a direct current (DC) filter or a high pass filter is disposed between noise and sampling circuits so as to remove an average value that deviates from a central value.
Korean Patent Application Publication No. 2007-0036799 discloses a method for sampling a signal, analyzing the sample value using a signal processing circuit, and then feeding back an offset controlling signal to a sampling circuit with an offset function for input signal decision threshold.
Examples of other methods using signal processing include a method for scrambling sample signals by a signal processing circuit disposed after a sampling circuit (Korean Patent Application Publication No. 10-2005-0084153) and a method for analyzing and equalizing sample values (Korean Patent Application Publication No. 10-2003-0027724).
However, the inventor(s) experienced that among these methods for overcoming the biasing, the method for feeding back deviation of a signal source possibly causes an unintended DC deviation. The inventor(s) also experienced that In the signal processing method such as scrambling, equalization, and so on, when using an algorithm, the merit of a generated random number as a random number is degraded due to its calculability like a pseudo random number, and a relatively complex or high-performance circuit is needed, requiring a long time for signal processing such as scrambling, equalization, and so on.